Research map / Detection and segmentation / Open-set and few-shot detection
GroupViT: Semantic Segmentation Emerges From Text Supervision
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add hierarchical GroupViT grouping instead of non-parametric grouping of ViT features | ViT text-supervised segmentation · PASCAL VOC 2012 | mIoU: 25.0 -> 52.3 | Helped |
| Add hard assignment instead of soft assignment (ablation) | GroupViT · PASCAL VOC 2012 val | mask mIoU: 12.0 -> 36.7 | Helped |
| Add multi-label contrastive loss (ablation) | GroupViT with hard assignment · PASCAL VOC 2012 val | mask mIoU: 36.7 -> 39.3 | Helped |
| Increase group tokens from 16 to 32 while holding output tokens at 8 (ablation) | GroupViT · PASCAL VOC 2012 val | mask mIoU: 37.1 -> 38.3 | Helped |
| Increase group tokens from 32 to 64 while holding output tokens at 8 (ablation) | GroupViT · PASCAL VOC 2012 val | mask mIoU: 38.3 -> 39.3 | Helped |
| Increase output tokens from 4 to 8 with 16 group tokens (ablation) | GroupViT · PASCAL VOC 2012 val | mask mIoU: 28.6 -> 37.1 | Helped |
| Increase output tokens from 8 to 16 with 64 group tokens (ablation) | GroupViT · PASCAL VOC 2012 val | mask mIoU: 39.3 -> 38.0 | Hurt |
| Add a second grouping stage (ablation) | GroupViT trained on CC12M · PASCAL VOC 2012 val | mask mIoU: 39.3 -> 41.1 | Helped |
| Add a second grouping stage (ablation) | GroupViT trained on CC12M+YFCC · PASCAL VOC 2012 val | mask mIoU: 37.2 -> 52.3 | Helped |
| Replace YFCC with RedCaps while retaining CC (ablation) | GroupViT · PASCAL VOC | mIoU: 52.3 -> 50.8 | Mixed |
| Replace YFCC with RedCaps while retaining CC (ablation) | GroupViT · PASCAL Context | mIoU: 22.4 -> 23.7 | Helped |
| Replace YFCC with RedCaps while retaining CC (ablation) | GroupViT · COCO | mIoU: 24.3 -> 27.5 | Helped |
Closest papers by meaning
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.89)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.88)
- GroupContrast: Semantic-aware Self-supervised Representation Learning for 3D Understanding (CVPR 2024 · similarity 0.87)
- Native Segmentation Vision Transformers (NeurIPS 2025 · similarity 0.87)
- Seg4Diff: Unveiling Open-Vocabulary Semantic Segmentation in Text-to-Image Diffusion Transformers (NeurIPS 2025 · similarity 0.87)
- Perceptual Group Tokenizer: Building Perception with Iterative Grouping (ICLR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.